On 13.02.2011, at 10:39, David Gibson wrote:
> On Sat, Feb 12, 2011 at 05:09:39PM +0100, Alexander Graf wrote:
>> On 12.02.2011, at 15:54, David Gibson wrote:
> [snip]
>>> + /* Don't generate spurious events */
>>> + if ((cur_level == 1 && level == 0) || (cur_level == 0 && level != 0)) {
>>
>> Did you hit this? Qemu's irq framework should already ensure that
>> property. I'm also not sure it's actually correct - if a level
>> interrupt is on, the guest would get another interrupt injected, no?
>> That would be cur_level ==1 && level == 1 IIUC.
>
> [snip]
>>> + case POWER7_INPUT_CKSTP:
>>
>> POWER7 has checkstop?
>
> [snip]
>>> + case POWER7_INPUT_HRESET:
>>
>> Does this ever get triggered? POWER7 is run in lpar only, so there is no
>> hreset, right?
>
> [snip]
>>> + case POWER7_INPUT_TBEN:
>>> + LOG_IRQ("%s: set the TBEN state to %d\n", __func__,
>>> + level);
>>> + /* XXX: TODO */
>>
>> Hrm - what is this?
>
> Ah, drat. I forgot about this. The POWER7 interrupt stuff I copied
> from 970 and them modified minimally to get it working. I meant to
> get around to auditing this stuff to see what was actually relevant to
> POWER7. I'll address this for the next version.
>
> [snip]
>>> +#if !defined(CONFIG_USER_ONLY)
>>> + env->slb_nr = 32;
>>
>> POWER7 has 64, no? Please check this :).
>
> Nope. POWER4 and POWER5 have 64, but POWER7 has 32. This one I did
> check and change.
Oh? Interesting. Good to know :)
Alex
